Postproc error

Hi,

I set up my suite (u-ch249) to run for 1 year, with both postproc and pptransfer turned on, so the data will get shifted over to JASMIN. This worked fine for the first 3 months of the model run, however the postproc step fails on month four (it is stuck on the ‘retrying’ stage, so maybe it didn’t totally fail, I’m not sure). I initially though this was because my work directory on archer2 was full, so I deleted the archived data for the first 3 months, however it is still failing with this error. Would you know how to fix this?

Liam.

Liam

It looks like something went wrong converting a fields file to pp format:

grenvill@uan01:/home/n02/n02/lpb20/cylc-run/u-ch249/share/data/History_Data> ls -l ch249a.p919790401*
-rw-r--r-- 1 lpb20 n02 198049792 Oct 12 09:30 ch249a.p919790401
-rw-r--r-- 1 lpb20 n02         0 Oct 12 07:57 ch249a.p919790401.pp

I’m not sure why um-convpp failed. Please try this:

cp /home/n02/n02/grenvill/ch249a.p919790401.pp /home/n02/n02/lpb20/cylc-run/u-ch249/share/data/History_Data/ch249a.p919790401.pp

then retrigger the post-proc task for 19790401T0000Z

Grenville

Liam

On further reflection - this may not be the right answer (it won’t harm to try.)

Grenville

Hi Grenville,

Yeah, I went ahead and made those changes, retriggered postproc but got back the exact same error. Is there any other reasons you can think for the error?

Liam

Liam

The real problem has arisen because the dump time is longer than the cycle time - the model log file points out

U_MODEL_4A: Warning: exiting at a period that is not a dump period
Therefore continuing the run will rerun preceding timesteps
This is inefficient and can cause restart problems

And indeed it has:
The first cycle/month it runs to the end of 30 days - no dump
Second cycle/month: restarts from beg of month 1 again and runs to end of month 2 - still no dump
Third cycle/month: starts from beg of month 1 and runs to the end of month 3 and dumps.
Fourth cycle/month: starts from beg of month 4 using last dump runs to end of month 4 - no further dump
So while postproc for fourth cycle is running the fifth cycle is running the model again from month 4 and thus overwriting the files that the postproc in the previous cycle is trying to process, etc!

This accounts for the strange time stamps on the fields files and the associated pp files.

I think the entire run is suspect. I’d start again, and set the dump time to 30 days (it’s currently set to 90 days)

Grenville

Hi Grenville,

Thanks, yeah, I think the dumping time must have been the issue, I set it to 30 days and it ran fine, and the output files make more sense as well. Thanks for the help.

All the best,
Liam

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.